RunDiffusion Photo

The worlds most photorealistic SDXL model is here!

Embrace unparalleled photorealism in our newest Stable Diffusion model to date. Partner with us to gain access to our stunning model, which will breathe life into your existing Stable Diffusion workflows. Commercial Licensing options available.

Try it!

Click to view in high resolution!

RunDiffusion Photo?

The RunDiffusion Photo Model is insane. If you haven't seen the examples above, stop now and view them. All done? Great!

Throw away all your weird tokens, "4k, perfect hands, detailed skin, HDR, RAW, photograph, flawless, etc". Just get rid of them! They're not needed. Take a look at the generation parameters that made this photo.

RunDiffusion Photo again brings ease, versatility, and beautiful image generation to your doorstep.

Prompt: beautiful woman wearing sneakers, full body shot.

Steps: 40, Sampler: DPM++, CFG scale: 5, Size: 832x1216, Model: RunDiffusion-Photo.safetensors

Dial it Up a Notch

In the hands of proficient users, RunDiffusion Photo becomes a formidable tool. Feel free to be descriptive and allow the model to take control. There's no greater feeling than getting a generation back, sitting back in your chair and saying, "Perfect!"

Prompt: perfect mix between Carnage from Marvel and Godzilla, with Carnage's red tentacles protruding from his entire body, destroying city, blood, epic, (cataclysmic, white eyes like venom, gargantuan, taller than buildings, oversized, gigantic:1.9), absurdres, best quality, fantasy style ,(intricate details), (hyperdetailed), 8k hdr, high detailed, lot of details, high quality, soft cinematic light, dramatic atmosphere, atmospheric perspective

Steps: 40, Sampler: DPM++, CFG scale: 3, Seed: 3044932836, Size: 832x1216, Model: RunDiffusion-Photo.safetensors

Try it in RunDiffusion Fooocus!

If you want to give RunDiffusion Photo a spin, create an account in our app, launch Fooocus, select RunDiffusion XL Photo in the model dropdown (under "Advanced") and fire away. Recommended settings are to use a low CFG (3 - 4.5), no refiner, no additional LoRAs, deselect all the styles that Fooocus auto selects, then generate.